Notice the proper 50 dB vertical axis to account for myriad of unknowns and accuracy issues. Erin zooms that out almost 10X to just 6 dB vertical scale!!! If you see issues in Soundstage tests, then yes, it is a problem. And it would show just as readily in my sweeps as I showed earlier.
I take your point and agree to some extent. I think it's academically interesting to see zoomed in, but is the difference between 0.5 ond 0.8 dB compression audible? Doubtful. Do people bitch and moan about it in stupid ways? Sure. OTOH, there's a whole subculture here dedicated to bitching and moaning about irrelevant differences in SINAD in pointless-me-too-whatever commodity electronics products, so...
I just showed the differential in KEF R11 Meta:
I like that. Please keep doing it.
(Though a normalized difference graph a la Soundstage would be easier to read IMO.)
If nobody looked at the graphs it would be one thing but now I read people saying that graph is one of the most important things in Erin's testing! Talk about creating a non-problem and getting people concerned for no good reason.
The two points are mutually exclusive: one can have some quibble about the presentation, while thinking the data is very important and useful.
We are not seeing anything remotely like that (4 dB loss across three octaves). I am sure there are some way under-designed drivers out there that have issues like this. The topic here are Erin publishing measurements after measurements on every speaker claiming to show reduction or in some cases increase (!) in relative output.
First, not
every speaker. The ones with really really good drive units, and especially good big drive units, come out looking better in Erin's compression testing than speakers with typical home audio smallish diameter, small voicecoil, minimal thermal management drive units. So just based on that kind of smell test intuition there's some value to it.
Also, I disagree with your implication that showing increases in level demonstrates flaws in the testing method. Rather, looking
where levels increase disproportionately points to reasonable causal mechanisms, such as port issues/passive radiator nonlinearity, cabinet resonances, maybe crossover issues, etc.